JPH0657063B2 - Motion vector detection method - Google Patents

Description

TECHNICAL FIELD The present invention relates to a motion vector detection method in a high-definition MUSE method.

(Outline of the invention) In the high-definition MUSE system, the camera pans,
When tilting, the entire screen becomes a moving pixel and field interpolation is applied, so the entire screen is blurred.
In order to prevent this, the MUSE method detects the panning and tilting speeds on the encoder side and transmits them to the decoder side, changing the read positions of the frame memory and field memory by the amount of the speed, and interpolating in the time axis direction. Is applied to prevent deterioration of image quality. This is called a motion vector correction method.

On the other hand, if the camera is following a moving object and panning, the object is stationary on the screen,
The motion vector is occurring. When motion vector correction is performed in such a scene, the surrounding scenery becomes a locking pixel and the object becomes a motion pixel, and the object becomes blurred. Human vision has a high probability of looking at the object, and it is not preferable that the object is blurred.

The method of the present invention is a motion vector detection method capable of removing the unnaturalness of an image by prohibiting the conventional motion vector correction by panning, tilting or the like for such a scene.

(Prior Art) Japanese Patent Application Laid-Open No. 61-61868 is a conventional technology that can detect the panning and tilting speeds of a camera related to motion vector detection.
-269475 "Motion vector detection method", Japanese Patent Laid-Open No. 62-2558
There is No. 7 "Motion vector detection circuit".

Both of them are representative point type motion vector detection using block matching, but the former invention divides the television screen into a plurality of small blocks, detects the motion vector for each block, and then comprehensively detects them. In other words, when the motion vectors of the respective divided areas substantially match, it is judged to be pan / tilt and one motion vector is generated. In the latter invention, if there is a still pixel in the moving image, this pixel is removed from the motion vector detection. Even if a still pixel is detected from the moving screen, Corresponds to the present invention described in detail later, but is the opposite of the main point of the present invention in that a still pixel is excluded from motion vector detection.

(Problems to be Solved by the Invention) The above-mentioned Japanese Patent Laid-Open No. 62-25587 is a very effective means for a device using a motion vector-corrected image itself, for example, a system conversion device. It is not always effective in the high-definition MUSE method that is targeted in. The reason is that the image quality deterioration in the part that does not hit the motion vector is the beat due to the difference in field rate in the system conversion, that is, the jerkiness of the motion, whereas in the high-definition MUSE system it is just a blur. .

Jerkiness is more severe in human vision than blurring, and when the camera follows a small moving object, it is not preferable that jerkiness occurs in the background, which is a large area on the screen. On the other hand, in the high-definition MUSE method, from the viewpoint that human vision is focused on this moving small object, and the background of the small object that is moving on the camera work is already blurred due to the accumulation effect of the image pickup tube. Therefore, it is preferable that the moving small object is the main target of motion vector detection.

However, none of the conventional motion vector detecting methods including the techniques described in the above-mentioned two patent publications can satisfy the above-mentioned requirements. Therefore, the object of the present invention is to detect a stationary portion from the screen shot by a camera that follows a moving small object with a relatively simple circuit configuration, , A motion vector detection method that can prohibit the information of the motion vector to be transmitted to the decoder side and that satisfies the above-mentioned requirements is provided. To do.

(Means for Solving the Problems) In order to achieve this object, the motion vector detection method of the present invention is to detect a panning speed and a tilting speed of a camera on the encoder side and transmit them as a motion vector, and a decoder In the motion vector detection method applied to the motion vector corrected television signal encoder / decoder system, the read position of the frame or field memory is changed by the size of the motion vector transmitted from the input side. A predetermined number of pixels is selected in the center of the screen, the degree of the still image is tested based on a predetermined threshold value for each of the selected pixels, and the number of pixels determined to be a still image by the test is the predetermined number. Motion used in the encoder / decoder system when the number of pixels exceeds a predetermined percentage It is characterized in that so as to prohibit the output of the vector detection circuit.

(Operation) Since the main body of the method of the present invention is to prohibit the transmission of a normally detected motion vector to the decoder side at least when there are a large number of stationary pixels at least in the center of the screen. The embodiment according to the invention results in counting the number of pixels with a motion vector at zero in the central part of the screen.
The reason for counting at the center of the screen is that the object of follow panning is often present at the center of the screen.

(Embodiment) The present invention will be described in detail below with reference to the accompanying drawings with reference to embodiments, but before that, reference will be made to an algorithm according to the present invention.

The algorithm of the present invention is as follows. Now the second
As shown, the target pixel on the screen is X, and the pixels on the left, right, top, and bottom of X are A, B, C, and D as shown in the figure. The pixel delayed by one frame of X is set to Y, and the absolute value of the signal difference between Y and all pixels is calculated. The absolute values of these signal differences are defined by the same lowercase alphabet. That is, x = | X−Y | a = | A−Y | b = | B−Y | c = | C−Y | d = | D−Y | Here, X, Y, A, B, C and D also represent output signal values at those pixels. At this time, the algorithm of the present invention (1) compares the minimum value of the absolute values a, b, c, d with the absolute value x, and when the absolute value x is smaller, the pixel is determined to be a still pixel.

(2) Still pixels are counted in the center of the screen for a period of one field. If this count value is larger than a fixed threshold, it is determined that the center of the screen is a still image, and transmission of motion vector signals is prohibited. .

(3) However, if the absolute value x is not sufficiently small, for example, even if the condition of (1) above is satisfied, it is excluded from the counting of (2) above. Conversely speaking, the above (1) is satisfied and the absolute value x
Means counting only if is small enough.

(4) In order to avoid the influence of noise, when the absolute value x has a value very similar to the minimum value of the absolute values a, b, c, d, even if the above (1) is satisfied, the above (2) To be excluded from the counting.
This means that the counting is performed only when the absolute value x is extremely smaller than the absolute values a, b, c and d.

FIG. 1 shows a block diagram of the configuration of an embodiment which satisfies the system of the present invention. The input signal is a luminance signal, which is guided to the motion vector detection circuit 1 and the 3-stage shift register 2. The motion vector detecting circuit 1 is disclosed in, for example, Japanese Patent Laid-Open No. 61-269475.
The examples described in the specification of No. 62-25587 may be used as they are. The three-stage shift register 2 operates at the clock rate of the input signal, and its output is guided to the latch circuit 3.
It is not necessary to take all pixels as the intervals of the target pixel X on the screen, and even if the intervals are thinned to a considerably coarse interval, for example, an interval of 4 to 16 points, statistically probable difference will not occur. However, X requires several thousand points for each field. As a result, the signal processing speed of the circuits after the latch circuit 3 becomes low, and the EP-ROM
Not only can an inexpensive element such as (erasable read-only memory) be used, but the capacity of the frame memory as the frame delay circuit 5, which will be described later, can be small, and the hardware is extremely simple.

The latch circuit 3 takes out three consecutive pixels in the horizontal direction, and delays these by 1H or 2H by a group of 1H delay circuits 4 (H is the interval of the horizontal synchronizing signal). Pixel signals corresponding to C, D and X are obtained. The signal Y is obtained by delaying the signal X by one frame by the frame delay circuit 5. The absolute value of the difference between the signal Y and the signals A, B, C, D and the signal X is taken by the group of subtractors 6 and the group of absolute values 7 to obtain the absolute values a, b, c, d, x. After that, the operation is as described in the above algorithm.

That is, the minimum value selection circuit 8 selects the minimum value among the absolute values a, b, c, d, and the amplitude limiter 9 limits the amplitude. A certain amplitude limit value is defined, and if it is less than this value, it remains as it is. This is equivalent to the condition (3) of the above algorithm. That is, the absolute value x must be smaller than this amplitude limit value for counting. It is convenient to use an EP-ROM as the amplitude limiting circuit 9. The weighting circuit 10 weights the absolute value X. This is the condition (4) of the above algorithm
This output z is, for example, z = kx + 1. However, k is preferably larger than 1 and smaller than 2, and l is a noise level equivalent value. That is, when the absolute value x has a value similar to the minimum value of the absolute values a, b, c, d by making a difference from the minimum value of the absolute values a, b, c, d by the weight, It can be prevented from being counted. The weight adding circuit 10 can also be easily realized by using an EP-ROM.

Subsequently, the comparator 11 compares the amplitude-limited minimum values a, b, c, and d with the weighted absolute value x, and when the weighted absolute value x is smaller, It is determined that this pixel is a pixel having a motion vector of zero, and the AND circuit 13 and a counter are used to AND this in the central portion of the screen (40% of the screen is appropriate) by the output of the mud generation circuit 12.
Lead to 14 and count over one field period. The result of this counting is the fixed threshold generator 15 and comparator.
When the fixed threshold defined by the generator 15 by 16 is exceeded, the output of the motion vector detection circuit 1 is prohibited by the prohibition gate circuit 17. That is, when the above conditions are satisfied, the output of the motion vector signal detected by the motion vector detection circuit 1 is prohibited so as not to be output.

In another embodiment of the present invention, in FIG. 2, four points in the diagonal direction may be added in addition to the front, rear, left and right of the target pixel,
On the contrary, considering that the follow panning is often performed in the horizontal direction, it can be simplified to only the right and left of the target pixel. In this case, the clock for thinning out the target pixel is rotated by a predetermined number of lines for each line, as shown in the above-mentioned Japanese Patent Laid-Open No. 25587/1987, for example, as the simplest example, it is inverted for each line. If so, the detection effect can be further increased. However, it will be apparent to those skilled in the art that the configuration of the embodiment of the present invention is not limited to that shown in FIG.

(Effects of the Invention) As described above, the present invention is contrary to the fact that the number of pixels that can be three-dimensionally interpolated (temporal axis interpolated) increases when the motion vector is corrected. It emphasizes the visual characteristics of where the human vision is likely to see. In other words, when the television camera is following a moving object, the 3D interpolation is applied to the moving object instead of the surrounding scenery from the viewpoint that human vision is likely to see the moving object. It is a technical idea to prevent blurring,
As a result of applying it to the MUSE method, good results were obtained as expected for the above scenes.

[Brief description of drawings]

FIG. 1 is a block diagram of an embodiment of the system of the present invention, and FIG. 2 is a diagram for explaining an algorithm of the system of the present invention. 1 ... Motion vector detection circuit 2 ... 3-stage shift register 3 ... Latch circuit, 4 ... 1H delay circuit 5 ... Frame delay circuit, 6 ... Subtractor 7 ... Absolute value unit, 8 ... Minimum value selection circuit 9 ... Amplitude limiter, 10 ... Weighting circuit 11, 16 ... Comparator, 12 ... Mad generation circuit 13 ... AND circuit, 14 ... Counter 15 ... Fixed threshold generator 17 ... Inhibition gate circuit

Claims (1)

[Claims] 1. The panning speed of the camera on the encoder side,
A television signal encoder / decoder in which the tilting speed is detected and transmitted as a motion vector, and the read position of the frame or field memory is changed by the size of the motion vector transmitted on the decoder side to correct the motion vector. In the motion vector detection method applied to the system, a predetermined number of pixels is selected from the input television signal in the central part of the screen, and the degree of the still image is tested based on a predetermined threshold value for each selected pixel. When the number of pixels determined to be a still image by the test exceeds a predetermined percentage of the predetermined number of pixels, the output of the motion vector detection circuit used in the encoder / decoder system is prohibited. A motion vector detection method characterized by the above.